Heying, Heather, 2001. Chapter 2 Social and Reproductive Behavior in the Madagascan Poison Frog, Mantella laevigata, with Comparisons to the Dendrobatids Abstract I present the first behavioral study of natural populations of a Madagascan poison frog. Focal watches of marked individuals were conducted for 925 hours, in five populations, across two seasons. Like the New World dendrobatids, these diurnal anurans eat ants and are aposematically colored. Data are presented that provide additional instances of convergence with the dendrobatids, including 1) extended male-male fights over defended resources necessary for reproductive success of both sexes, 2) stereotyped, highly tactile courtships in which the female may reject initial oviposition sites, and 3) complex maternal care. Females return to water-filled phytotelmata, or wells, and lay trophic eggs for their tadpoles. Mantella laevigata has the minimum possible clutch size in anurans— usually one—suggesting a high degree of parental investment. Males defend wells, which attract females who oviposit in the wells. Fertilized eggs may hatch and metamorphose, or may be eaten by a tadpole already in that well, of which the territorial male is probably the father. Unfertilized eggs serve as food for tadpoles. Oviposition-site scouting behavior of both sexes, and the dependency of tadpole presence on the position of eggs laid, provide evidence of 38 Chapter 2 of dissertation in Biology, University of Michigan Heying, Heather, 2001. context-dependent, and assessment, behavior. Females leave courtships most often only after visiting potential oviposition sites, while males usually leave to engage other males in aggression, suggesting that territory maintenance may be the most important component of male reproductive success. Two other species of frogs often prevent M. laevigata from using defended oviposition sites, and larval crane-flies predate the eggs of all frog species using water-filled wells. Introduction The poison frogs of Madagascar, genus Mantella, have long been of interest to students of anuran behavior, largely because there are many similarities between species of Mantella and the toxic members of the neotropical family Dendrobatidae, the poison-dart frogs. Many species of dendrobatids, especially those in the genera Dendrobates (sensu lato) and Phyllobates, are known for their territoriality, stereotyped courtship sequences, and complex parental care. Until now, nothing was known of the social system of any species of Mantella. This study marks the first behavioral research conducted on Mantella in the field. When Boulenger identified Mantella in 1882, he placed the genus in Dendrobatidae. Although he later formalized this taxonomic placement on the basis of their shared lack of teeth (Boulenger 1914), it is clear that he felt that the two groups simply shared a gestalt (Boulenger 1882). Though Boulenger did not know it, the aposematic coloration in both groups arises, in part, from identical classes of lipophilic alkaloids in the skin, including pumiliotoxins and 39 Chapter 2 of dissertation in Biology, University of Michigan Heying, Heather, 2001. decahydroquinolines (Daly et al. 1996). Among anurans, these classes of toxins are otherwise known only in two other (unrelated) frog genera [Melanophryniscus (Bufonidae) and Pseudophryne (Myobatrachidae), Daly et al. 1984]. Other similarities between Mantella and the dendrobatids include small size (most species have a snout-vent length of 20-30 mm), diurnality, terrestrial and/or arboreal habit, and a diet consisting primarily of ants and mites. Captive dendrobatids lose their toxicity (Daly et al. 1980; Daly et al. 1992), as do captive Mantella (Daly et al. 1997), suggesting that the building blocks for the alkaloids are dietary. A high percentage of ants in the diet of dendrobatids is correlated with toxicity (Caldwell 1996), and some ants (genus Solenopsis) in the dendrobatid diet produce some of the simpler alkaloids themselves (Spande et al. 1998; Jones et al. 1999). Mantella laevigata, commonly known as the climbing Mantella, has more characters in common with the aposematic dendrobatids than do the other ten species of Mantella. It is not fully terrestrial, and has expanded toe-pads. It also breeds in water-filled tree holes or broken bamboo (phytotelmata), much like the preferred bromeliads of some dendrobatids. Amplexus is neither reliably axillary nor inguinal in M. laevigata, which are the two most common mating positions in anurans; M. laevigata adopts a variety of positions during mating amplexus (pers. obs.). In several dendrobatids, amplexus is also unlike that of most anurans, being absent (Crump 1972; Limerick 1980), or cephalic, a highly derived form in which the male grasps the female by the head (Myers et al. 1978; Wells 1980). Finally, the larval morphology of M. laevigata resembles that of the oophagous dendrobatid tadpoles, having reduced labial denticles and a large, keratinized horny beak (Glaw & Vences 1994). 40 Chapter 2 of dissertation in Biology, University of Michigan Heying, Heather, 2001. Several other taxa are known from both South America and Madagascar, but not Africa (e.g., side-neck turtles, Underwood 1976; iguanians, Frost & Etheridge 1989; boas, Kluge 1991). These examples are consistent with reconstructions of Gondwana that suggest that Africa and South America partially broke from the rest of the Southern landmass by the late Jurassic, and that by the early Cretaceous, Africa and South America were rifting, leaving a more recent and direct land connection between South America and Madagascar than between Africa and Madagascar (Scotese & Golonka 1992). These observations might lead to a hypothesis of sister-group relationship between the dendrobatids and Mantella, but phylogenetic analyses, though not complete with regard to the taxa investigated, provide evidence against such a historical hypothesis (e.g. Hay et al. 1995). Thus, it seems that all similar specializations between the dendrobatids and Mantella are independently evolved. While a dendrobatid-Mantella relationship would support the hypothesis of Gondwana reconstruction outlined above, it provides little insight into more general causal chains that may have led to what we observe today. Complex convergences, involving distinct processes, can help identify environmental factors that might lead to similar results in the different organisms. That is, the known existence of several convergences may suggest that others are present. Based on the known similarities described above, my goal was to discover the social system and natural history of M. laevigata, with the broad hypothesis that some or all of the behaviors of aposematic dendrobatids would be present in M. laevigata. I predicted that male M. laevigata would be territorial, defending calling and/or oviposition sites from other males, as several dendrobatids do (e.g., D. granuliferus, Crump 1972; D. pumilio, Bunnell 1973; E. femoralis, 41 Chapter 2 of dissertation in Biology, University of Michigan Heying, Heather, 2001. Roithmair 1992). I predicted that M. laevigata would have a stereotyped courtship sequence similar to that of some dendrobatids, in which a male attracts a female using advertisement calls, changes to a softer courtship call, and through repeated tactile interaction with the female leads her to one or more potential oviposition sites until she accepts one, or abandons the courtship (see review by Wells 1977; also Colostethus stepheni and C. marchesianus, Juncá 1998). I also predicted that M. laevigata would have complex maternal care of the type found in the oophagous dendrobatids, in which mothers return to individual tadpoles repeatedly to feed them trophic (unfertilized) eggs (see reviews by Weygoldt 1987; Crump 1996). This behavior has also been observed in three unrelated species of anurans, two hylids (Osteopilus brunneus, Lannoo et al. 1986; and Anotheca spinosa, Jungfer 1996; Thompson 1996), and one rhacophorid (Chirixalus eiffingeri, Ueda 1986; Kam et al. 1996). In my 1999 field season, I had already failed to falsify these broadly defined "hypotheses of convergence,” based on the behavior I had observed in M. laevigata in 1997. I then focused on quantifying what variables affect the outcomes of territorial encounters and courtships, and what and how often other organisms in the phytotelmata (well) community interact with M. laevigata. With these data it is possible to construct a coherent picture of the selective pressures affecting reproductive success for these frogs. 42 Chapter 2 of dissertation in Biology, University of Michigan Heying, Heather, 2001. Methods Mantella laevigata was observed during the Malagasy summers of 1997 and 1999 (January through May, and January through April, respectively). The research was conducted on the 510 hectare island of Nosy Mangabe, which lies at the northern end of the Bay of Antongil, five kilometers south of the town of Maroantsetra, Toamasina Province, in northeastern Madagascar (15º 30’ S, 49º 46’ E). Nosy Mangabe is a “Special Reserve” in the Masoala National Park consisting mainly of 100 – 400 year old secondary forest. There are two summits, the highest being 332 m, and most of the island is inaccessible by trail. M. laevigata are found throughout the island, but are densest in coastal bamboo stands. The bamboo on Nosy Mangabe--Bambusa vulgaris--was introduced, but in Marojezy, the other region in which M. laevigata is found, the frogs are found in association with native bamboo species,